Build your first workflow

Module 3. Your goal: Build a three-node image workflow and explain how the image moves through it. You will create an image with Seedream, save the result and keep a workflow you can reuse.

Before you start

Complete Module 1 and Module 2.
Confirm access to Seedream 5.0 Pro in your key’s region.
Have the successful first image and product brief from Module 2.
Use the practice logo for the walkthrough, or upload a reference image you have permission to use.

Download the practice logo (PNG). Keep this reference file with your workflow; it is not included inside the JSON.

Keep your key in Settings > BytePlus. Never put it in a prompt, workflow or screenshot. Running generation uses your BytePlus account balance.

What you will build

Load Image -> Seedream -> Save Image

A node performs one step. A connection passes data between nodes. A workflow is the connected graph, including its settings.

Load Image supplies your reference through its IMAGE output.
BytePlus Seedream combines that reference with your prompt and generates an image.
Save Image receives the generated image and writes it to ComfyUI’s output folder.

Load Image connected to the first Seedream reference input, with Seedream's IMAGE output connected to Save Image — Follow the two image connections from left to right.

The screenshot may show a model-version label in the node title. Search for BytePlus Seedream in the current pack.

The included Seedream template is your worked example. Open it from Templates > ComfyUI-BytePlus-ModelArk if you want to compare the connections. For this lesson, build the graph yourself.

Template: Seedream, in the ComfyUI template browser under BytePlus
BytePlus nodes: Seedream 4.5 & 5.0
Keys needed: ModelArk API key
Also uses: Load Image, Save Image (built into ComfyUI)

Build the workflow

Press + beside the workflow tabs to start a blank workflow.
Open Nodes in the left sidebar. Search for Load Image, then drag its result onto the canvas. Repeat for BytePlus Seedream from the BytePlus ModelArk category, and Save Image.
Move each node by dragging its title bar. Arrange Load Image, Seedream and Save Image from left to right.
In Seedream, choose seedream 5.0 pro. Press and hold the small IMAGE circle on the right of Load Image, drag to Seedream’s image_1 circle on the left, then release. A line should remain.
Drag Seedream’s IMAGE circle on the right to Save Image’s images circle on the left. Leave MASK, mask and response unconnected.
In Load Image, press choose file to upload and select the downloaded logo or your reference. Confirm a preview appears.
Select (2K) 2048x2048 (1:1) for size_preset. For the logo example, paste the prompt below into Seedream. Keep other model settings at their defaults.

Nodes library with BytePlus Seedream searched and the generation node identified for dragging onto the canvas — Add the generation node, not Layer Separation.

Close-up showing Load Image IMAGE linked to Seedream image_1 and Seedream IMAGE linked to Save Image images — Drag from an output circle on the right to the matching input circle on the left.

If labels are too small, zoom in on the canvas. Inputs may show a prefix such as model.images.image_1; use the first reference image socket ending in image_1.

Turn this logo into a glossy 3D badge on a clean white studio background with soft shadows.

For your own reference, replace “this logo” with the subject and describe what you want to keep and change.

Before running, trace the graph: the reference enters Seedream, the prompt describes the change, and the generated image goes to Save Image. Image sockets connect to image sockets; the text response output is not an image.

Run and review

Press Run and wait for the Seedream node to finish.
Review the image displayed in Save Image. For the logo prompt, look for a 3D badge, a white background and soft shadows.
Find the saved image in ComfyUI’s output folder. The Save Image filename_prefix controls its name and subfolder: Course/product-image saves in output/Course, with filenames beginning product-image. ComfyUI adds a counter and file extension.

Example output for the logo prompt; your result can differ.

The generation node does not save images itself. Save Image both keeps the result and gives the graph an output to execute.

Copy the generated image to your course folder. The output folder is inside the ComfyUI file location from Module 1, not inside the folder where your browser downloads JSON.

Save and reopen

Open Graph in the top toolbar and choose Export, not Export (API).
Enter first-seedream-workflow and press Confirm. Locate the downloaded first-seedream-workflow.json in your browser’s download location and move it to your course folder.
Press + to open a blank workflow, then drag the saved JSON onto the canvas.
Confirm the three nodes, two connections and prompt are restored. Reselect or upload the reference image if it is not available on this machine.
Run the reopened image workflow and confirm an image appears in Save Image.

Graph menu identifying Export as the reusable workflow JSON option and distinguishing it from Export API — Export downloads your workflow; it is different from saving the generated image.

Export Workflow dialog with first-seedream-workflow entered as the filename above the Confirm button — Confirm the name, then collect the JSON from your download location.

The JSON stores the graph and settings, not the reference image file. Keep that image with your workflow when sharing. The recipient also needs ComfyUI, this node pack, model access and their own ModelArk API key.

You have now reopened your own saved workflow. Keep this export so you can reuse and share the graph you built.

Practice

Adapt the graph using the original product reference and brief you prepared in Module 2. Replace the reference and prompt; do not create another brief yet. If you chose the logo-badge practice, use that subject instead.

Try this starting prompt:

Keep the product's shape, colour and label from the reference. Place it on a clean white studio background with soft shadows. Do not add extra text or objects.

Review the result against the brief and note any unwanted changes to the product or label. Save the image and workflow, then explain what each node contributes.

Completion check

You built the graph yourself and can trace both image connections.
You can identify where to change the reference image and prompt.
A generated image appears in Save Image and is saved in output.
Your workflow JSON reopens with the same graph and settings.
Your shared materials include the reference image but no credentials.

Next lesson

Module 4: Create and refine product images.

Previous: Module 2: Set up your API key.

Return to the course learning path.